How significant are the known collision and element distinctness quantum algorithms?

Abstract

Quantum search is a technique for searching N possibilities in only O(sqrt(N)) steps. It has been applied in the design of quantum algorithms for several structured problems. Many of these algorithms require significant amount of quantum hardware. In this paper we observe that if an algorithm requires O(P) hardware, it should be considered significant if and only if it produces a speedup of at least O(sqrt(P)) over a simple quantum search algorithm. This is because a speedup of $O(sqrt(P)) $ can be trivially obtained by dividing the search space into $O(P)$ separate parts and handing the problem to independent processors that do a quantum search. We argue that the known algorithms for collision and element distinctness fail to be non-trivial in this sense.